COMPARING GOOGLE TRANSLATE AND INSTAGRAM

ENGLISH TRANSLATIONS OF CAPTIONS

Dwi Nur Asiyah Ningrum1, Issy Yuliasri2
English Education Department, Faculty Language and Arts
Universitas Negeri Semarang, Semarang, Indonesia
Email : dwinurasiyah560@gmail.com
Abstract : Nowadays, machine translators like Google Translate are widely used by people.
However, it does not mean that machine translators can always be relied on and
produce good translations. This study aimed to analyze the translation quality of
Instagram Machine Translation and Google Translate in terms of accuracy,
acceptability, and readability. The objective of this research was to find out which
machine translator is better between Instagram Machine Translation and Google
Translate in translating Instagram captions. The data from the questionnaire were
described in qualitative form. Nababan's translation quality instrument (2012) was
adopted to assess the quality of the translation results from Google Translate and
Instagram Machine Translation, where the translations were assessed with points
one, two, or three according to the quality of the translation. The results of the
analysis show that in three aspects, namely accuracy, acceptability, and readability,
Google Translate is superior to Instagram Machine Translation.

Keywords : Translation Quality, Instagram Translation, Google Translate

INTRODUCTION
Translation is quite a problem for many
people because it involves a person's
language skills. Although there is English,
which is used as the international language,
there are still many people, including
Indonesians, who do not master it. Based
on the data on EF EPI (English Proficiency
Index) web, Indonesia is categorized as
low-proficiency in English and ranked 15th
out of 24 Asian countries, while out of 100
countries in the world Indonesia is in
position 74th. The existence of a machine
translator such as Google Translate,
Microsoft Translator, iTranslate, Bing
Translation provides a solution to that
problem. A machine translator will provide
the core meaning of a text. Therefore, even
if it is considered inaccurate, the reader will
still understand the whole information of
the translated text (Meilasari,2019).
Instagram, a well-known social media and
widely used by the young generation, also
has an automatic translator feature that can
be used to translate captions, bios, and also
JELLT (Journal of English Language and Literature Teaching)
Vol. 7, No. 2 November 2022
P.ISSN : 2548-7728 E.ISSN : 2599-0373
comments.
Despite its existence which solves the
translation problem, machine translators
have drawbacks and give disadvantages to
their users if they are not used properly and
without human intervention. As revealed by
Peng (2018) in his journal, machine
translators have poor sentence construction
and word selection. Words and sentences
that have special meanings are also difficult
to translate, and often the result of machine
translation looks more like the source text
than the intended meaning. Thrivini (2002)
argues that social understanding and
knowledge of the characteristics of a
translation (the characteristics of a
translated text) must be considered so that
the translation can be completed properly.
This gives an understanding that in writing
a translation, a text containing cultural
elements cannot be translated by a machine
translator. Tripathi and Sarkhel (2010) say
that the existence of a machine translator is
used to assist and simplify the translation
process, but not to translate the entire
349
contents of the document without editing
because machine translators cannot provide
perfect translation results. The existing
machine translation applications may be
able to translate literal translations but are
not able to create an equivalent translation
between the source language and the target
language.
A study to compare human translators
with machine translators conducted by
Ahrenberg (2017) found that there are
several steps taken by human translators to
achieve a translated text that has a high
equivalence between the source text and the
target text. These methods cannot be done
by machine translators. Some of the most
important discoveries are sentence splitting,
sunction and category changes,
explicitation, and paraphrase. This is in line
with what was conveyed by Machen, Prou,
and Tezcan (2020) which says that
although machine translators have
developed, they still have not produced
consistent translations so that the
translation results cannot be published
immediately. To get high-quality
translations, human intervention is still
needed, which is called post-editing.
Many studies discuss the quality of a
machine translator, one of them is research
conducted by Halimah (2018). The purpose
of her study is to determine whether
Google translate can produce good
translations and to what extent translations
made by Google Translation are acceptable.
Content Analysis approach was applied in
this study to analyze English procedural
text on a "VIXAL Lebih Wangi". She
found that there were several differences
between the translation produced by
humans and the translation produced by
Google translate. The similarity between
the two translations is only 29% because
the choice of words used is not the same,
resulting in a translation that has a different
meaning. Human translation is more
acceptable than the translation produced by
Google translate so that in its use it requires
the role of humans to be able to produce
JELLT (Journal of English Language and Literature Teaching)
Vol. 7, No. 2 November 2022
P.ISSN : 2548-7728 E.ISSN : 2599-0373
good translations.
Another is research conducted by
Nainggolan, Sinurat and Hutauruk (2020)
focus on knowing the level of accuracy of
the translation results generated by Google
Translate. The subject of that research is
Rysava's (2012) article with the title “Noun
Phrase in English”. From 25 data of noun
phrases, the researcher found 11 datum of
noun phrases as accurate translations, 13
datum of noun phrases as less accurate
translations, and 1 datum of noun phrases as
an inaccurate translation. In conclusion, the
results of the Google translate translation in
translating a noun phrase are categorized as
less accurate.
Different from previous studies, this
research compared two machine
translations, namely Google Translate and
Instagram Machine Translation. The reason
why the researcher chose Instagram
Machine Translation is that Instagram is a
very popular social media and is widely
used by Indonesian people. While Google
Translate is a machine translator that is
known by many people so it will be
interesting when the two machine
translators are compared for their quality.
This research is expected to provide
benefits for Google Translate and Instagram
users so that they know the quality of the
two machine translators.
RESEARCH METHOD
Descriptive qualitative research was
used in this study. There were two types of
objects, including material objects and final
objects. For the material object in this
study, the researchers used Instagram
captions from accounts belonging to Leena
Snoubar (@withloveleena), and the final
object was a questionnaire assessed by
respondents using Nababan's theory of
translation quality (2012).
In collecting the data, there were several
stages carried out by researchers. The stages
included choosing some Instagram captions
from the @withloveleena account,
translating the selected captions using
350
Instagram Machine Translator and Google
Translate then writing down the translation
results, distributing the questionnaire, and
lastly, analyzing the results of the
questionnaire.
To analyze the data, the researchers
used an instrument proposed by Nababan.
There were three aspects assessed in this
instrument, namely accuracy, acceptability
and readability. The aspect of accuracy
emphasizes the equivalence of the content
or message in the text. Acceptability aspect
refers to whether the translation result has
been adjusted to the norms or customs that
apply in the target language or not. For the
readability aspect, according to what
Nababan said, it does not only refer to the
source language but also the target
language.
For the accuracy aspect, the
questionnaire was assessed by three
lecturers who have at least a master degree.
They were asked to give a score.
Translations that were given score 3 were
judged to be accurate, translation that were
given score 2 were judged to be less
accurate, and translation that were given
score 1 were judged to be inaccurate. In
acceptability and readability aspects, the
questionnaire was assessed by twenty
assessors who are native Indonesians. They
were asked to give a score with the range 1
to 3. Score 3 means that the translations
were acceptable for the acceptability aspect
and had a high level of readability for the
readability aspect, score 2 means that the
translation were less acceptable for the
acceptability aspect and had a medium
level of readability for the readability
aspect, and score 1 means that the
translation were unacceptable of the
acceptability aspect and had a low level
readability for readability aspect
FINDINGS AND DISCUSSION
After analyzing the data, it shows that
Google Translate has better translation
quality than Instagram Machine
Translation. This is because Google
Translate has a higher percentage of data
JELLT (Journal of English Language and Literature Teaching)
Vol. 7, No. 2 November 2022
P.ISSN : 2548-7728 E.ISSN : 2599-0373
that has 3 points in all aspects (accuracy,
acceptability, readability).
Accuracy of the Translation
The translation is considered accurate if
the meaning of the translation is translated
accurately and perfectly from source
language to target language without any
mistake. A score of two for translations
rated "Less Accurate". The translation is
rated as "Less Accurate" if the translation is
almost translated perfectly, but there is
distortion in meaning and it annoys the
message of the source language. The last
one is a score of one. A score of one is for
translations that are judged "Inaccurate"
which means that the translations are not
accurately translated into the target
language. There were 40 Instagram captions
used for this aspect of accuracy taken from
the Instagram account @withloveleena.
The questionnaire filled in by the
respondents shows that the translation from
Google translate has a higher percentage of
accurate data than the Instagram machine
translation. Google Translate has a
percentage of 32.5% for data that is
considered accurate or gets 3 points while
Instagram Machine Translation has a
percentage of 15%. Then for data that is
considered less accurate or gets 2 points,
google translate has a percentage of 58.33%
while Instagram machine translation has a
percentage of 61.61%. For data that is
considered inaccurate or gets 1 point,
google translate has a percentage of 9.17%
while instagram machine translation has a
percentage of 23.33%.
Table 1. The amount of data and
percentage on the aspect of accuracy
Quality
Accuracy
No Machine
(Inaccurate)
(Accurate)
Accurate)
. Translation
(Less
3
351
 Dear baby, you have a cute wardrobe

(61.67%)

(23.33%)
waiting for you if that makes you want

(15%)
1. IMT

18

74

28
to come any faster? � I have been
taking this week to rest, focus on taking
Total Data 120 Data care of myself and enjoying the end of
my pregnancy. As uncomfortable as I
am now, I started thinking about how

(58.33%)
(32.5%)

(9.17%)
2. GT 39

70

11
much I will miss her being in my
tummy yesterday and it made me so
Accurate emotional �
As explained above, the criteria for an Translation Caption :
item to be considered accurate are the word Sayang sayang, Anda memiliki lemari
meanings, technical terms, phrases, clauses, pakaian lucu yang menunggu Anda jika
sentences, or source language text itu membuat Anda ingin datang lebih
accurately transferred into the target cepat? � Saya telah mengambil minggu
language; absolutely no meaning ini untuk beristirahat, fokus merawat
distortions. diri sendiri dan menikmati akhir
Instagram Machine Translation kehamilan saya. Tidak nyaman seperti
saya sekarang, saya mulai berpikir
Instagram Machine Translator has a
tentang betapa saya akan
percentage of 15% for translations that are
merindukannya berada di perut saya
considered accurate or get 3 points. Below
kemarin dan itu membuat saya sangat
is an example of an accurate translation of
emosional �
Instagram Machine Translation:
On the result of the translation above,
(Datum A-18) all raters gave 3 points. It means that the
Original Caption : translation is accurate. Information from the
Some of my favorite past springtime source language translates well to the target
looks! �� which would you wear? language. Every word and sentence is
Translation Caption : translated precisely so that there is no
Beberapa penampilan musim semi distortion of meaning.
terakhir favorit saya! ��mana yang Less Accurate
akan kamu kenakan? Less Accurate data is data that most of
On the result of the translation above, the meaning of words, technical terms,
all raters gave 3 points. It means that the phrases, clauses, sentences, or source
translation is accurate. Information from language text has been transferred
the source language translates well to the accurately into the target language.
target language. Every word and sentence However, there are still meaning distortions
is translated precisely so that there is no or double-meaning translation (ambiguous)
distortion of meaning. or eliminated meaning, which interferes
with the integrity of the message.
Google Translate
Google Translate has a percentage for Instagram Machine Translation
accurate data or translation that gets 3 Instagram Machine Translation has a
points as much as 32.5%. Below is an percentage of 61.67% for translations that
example of an accurate translation of are less accurate or for translations that get
Instagram Machine Translation: 2 points. Below is an example of a less
accurate translation from Instagram
(Datum B-38)
Machine Translation:
Original Caption :
(Datum A-6)
JELLT (Journal of English Language and Literature Teaching) 352
Vol. 7, No. 2 November 2022
P.ISSN : 2548-7728 E.ISSN : 2599-0373
 Original Caption :
�� ��� Alaina and I are out of
office for the rest of the summer (I wish
�)
Translation Caption :
�� ��� Alaina dan aku keluar dari
kantor untuk sisa musim panas (aku
berharap �)
On this data, all raters give 2 points for
the translation results. This means that the
message from the source language is
translated well, but there are still meaning
distortions, ambiguous meanings or
missing meanings that reduce the
authenticity of the meaning.
Google Translate
Google Translate has a percentage of
58.33% on data that is considered less
accurate or gets 2 points. Below is an
example of a less accurate translation from
Google Translate:
(Datum B-3)
Original Caption :
Happy first Father’s Day to my husband
� I have loved watching you step so
naturally into the role of a father this
past year and couldn’t have asked for a
better man to lead our family!!! Alaina
is so blessed to have you as her Baba to
look up to! You hold us together and
are strong for us even when things feel
like they are falling apart � I love you
so much!!!
Translation Caption : not conduct interviews with the raters so the
Selamat Hari Ayah pertama untuk
suami saya � Saya senang melihat
Anda melangkah secara alami ke dalam
peran seorang ayah tahun lalu dan
tidak bisa meminta pria yang lebih baik
untuk memimpin keluarga kami!!!
Alaina sangat diberkati memiliki Anda
sebagai Baba yang harus dijunjung!
Kamu menyatukan kita dan kuat untuk
kita bahkan ketika hal-hal terasa
seperti berantakan � Aku sangat
mencintaimu!!!
Inaccurate
JELLT (Journal of English Language and Literature Teaching)
Vol. 7, No. 2 November 2022
P.ISSN : 2548-7728 E.ISSN : 2599-0373
Data is considered Inaccurate if the
word meanings, technical terms, phrases,
clauses, sentences, or source-language text
are not accurately transferred into the target
language, deliberately.
Instagram Machine Translation
Instagram Machine Translation has a
percentage of 23.33% for data that is
considered inaccurate or gets 1 point.
Below is an example of an inaccurate
translation from Instagram Machine
Translation:
(Datum A-19)
Original Caption :
Ventured 5 steps outside the front door
to document this winter wonderland!!!
We haven’t had power all day, but we
are cozied by the fire and enjoying
having nowhere to be ❄��
Translation Caption :
Berkelana 5 langkah di luar pintu
depan untuk mendokumentasikan dunia
dongeng musim dingin ini!!! Kami
belum memiliki kekuatan sepanjang
hari, tetapi kami dikejar oleh api dan
menikmati tidak ada tempat untuk
berada
All of the three raters gave one point for
the translation of the data. That means the
message from the source language is not
translated accurately to the target language.
However, in this study, the researcher did
researcher could not find out the reason
why two raters gave one score while one
gave two points.
Google Translate
Google Translate has a percentage of
9.17% for data that is inaccurate or gets 1
point. Below is an example of an inaccurate
translation from Google Translate:
(Datum B-38)
Original Caption :
Dear baby, you have a cute wardrobe
waiting for you if that makes you want
353
 to come any faster? � I have been
taking this week to rest, focus on taking
care of myself and enjoying the end of
my pregnancy. As uncomfortable as I
am now, I started thinking about how
much I will miss her being in my
tummy yesterday and it made me so
emotional �
Translation Caption : Machine Translation gets higher points, namely
Sayang sayang, Anda memiliki lemari
pakaian lucu yang menunggu Anda jika
itu membuat Anda ingin datang lebih
cepat? � Saya telah mengambil minggu
ini untuk beristirahat, fokus merawat
diri sendiri dan menikmati akhir
kehamilan saya. Tidak nyaman seperti
saya sekarang, saya mulai berpikir
tentang betapa saya akan
merindukannya berada di perut saya
kemarin dan itu membuat saya sangat
emosional �
Two of the three raters gave one point for
the translation of the data. That means the
translation that gets 3 points is 48% and on
Instagram, Machine Translation has as much as
22.62%. Then, for translations that get 2 points,
Instagram Machine Translation is superior to
Instagram Machine Translation. The data on
Instagram Machine Translation that gets 2
points is 51.5% while Google Translate is
44.5%. Furthermore, for translations that get
one point or are not acceptable, Instagram
as much as 25.88% while Google Translate is
only 7.5%.
Acceptable
Translation is considered acceptable if it
feels natural, the terms used are easy to
understand by the reader and the grammar
used is in accordance with the rules in the
target language.
Instagram Machine Translation
Instagram Machine Translation has a
percentage of 22.62% for translations that
are considered acceptable. Below is an
Quality
Acceptability
Machine (Acceptab

Acceptabl

Acceptabl
message from the source language is not No.
Translation

(Less

(Not
translated accurately to the target language.
le)

e)

e)
3

1
However, in this study, the researcher did not
conduct interviews with the raters so the
researcher could not find out the reason why
(22.62%)

(25.88%)
(51.5%)
two raters gave one score while one gave two
181

412

1. IMT 207
points.
Acceptability of the Translation
A translation is considered acceptable if the Total Data 800 Data
translations are natural; the technical terms are
commonly used and familiar to the reader;
(44.5%)
(48%)

7.5%
384

356

60

phrases, clauses, and sentences are in 2. GT

accordance with the rules of the Indonesian
language. Then a score of two is for the
translation that already feels natural; however,
there are a few problems in the use of technical
terms or grammatical errors. A score of one is
for unnatural translations; the technical terms
are not commonly used and not familiar to the
readers; phrases, clauses, and sentences are not
in accordance with the rules of the Indonesian
language.
The results of the questionnaire show that
Google Translate has a higher percentage of
accurate translations than Instagram Machine
Translation. On Google Translate the
JELLT (Journal of English Language and Literature Teaching)
Vol. 7, No. 2 November 2022
P.ISSN : 2548-7728 E.ISSN : 2599-0373
example of an acceptable translation from
Instagram Machine Translation:
(Datum A-4)
Original Caption :
Feeling so so grateful ��� I cannot
believe we only have a few more weeks
until we are a little family! God willing
�
Translation Caption :
Merasa sangat bersyukur ��� Aku
tidak percaya kita hanya memiliki
354
 beberapa minggu lagi sampai kita
menjadi keluarga kecil! InsyaAllah �
In the translation above, twelve raters
gave three points, six raters gave two points
and two raters gave one point. The
translation was categorized as acceptable
because there are half of the raters gave
three points. This indicates that the
translation is considered natural, the terms
used are familiar and the grammar used is
by the rules of the target language.
Google Translate
Google Translate has a percentage of
48% for data that is considered acceptable.
Below is an example of an acceptable
translation from Google Translate:
(Datum B-4)
Original Caption : kami! Menurutmu kita mirip atau tidak?
Feeling so so grateful ��� I cannot
believe we only have a few more weeks
until we are a little family! God willing
� raters who gave two points (14 raters)
Translation Caption : shows that the translation is quite natural
Merasa sangat bersyukur ��� Saya
tidak percaya kita hanya memiliki
beberapa minggu lagi sampai kita
menjadi keluarga kecil! Insya Allah �
In the translation above, twelve raters
gave three points, six raters gave two points
and two raters gave one point. The
translation was categorized as acceptable
because there are half of the raters gave
three points. This indicates that the
translation is considered natural, the terms
used are familiar and the grammar used is
by the rules of the target language.
Less Acceptable
Less Acceptable translations are
translations that already feel natural but
there are some phrases, clauses, or
sentences that are still not familiar to the
reader and are also translated not by the
grammar of the target language.
Instagram Machine Translation
Instagram Machine Translation has a
percentage of 51.5% for data that is
considered less acceptable. Below is an
JELLT (Journal of English Language and Literature Teaching)
Vol. 7, No. 2 November 2022
P.ISSN : 2548-7728 E.ISSN : 2599-0373
example of a less acceptable translation
from Instagram Machine Translation:
(Datum A-7)
Original Caption :
A rare photo of all of us together!
Everyone is making me look short
because they are wearing heels! Except
Naeem haha � We always get asked if
we are full siblings and we are! Do you
think we look alike or no?
Translation Caption :
Foto langka kita semua bersama!
Semua orang membuatku terlihat
pendek karena mereka mengenakan
sepatu hak tinggi! Kecuali Naeem
hahaha � Kami selalu ditanya apakah
kami saudara kandung lengkap dan
In this data, three raters gave three
points, fourteen raters gave two points and
three raters gave one point. The number of
but there are still clauses, phrases and
sentences that are less familiar to readers.
Google Translate
Google Translate has a percentage of
51.5% for data that is considered less
acceptable. Below is an example of a less
acceptable translation from Google
Translate:
(Datum B-3)
Original Caption :
37 weeks! � My app says the baby is
around 6.5 lb and the size of a small
melon! Baby is practicing her karate
moves while I practically stand on my
head trying to get her to turn into the
head down position �� My hospital
bag is packed and I’ve got all the gear
and no idea �� but we are so excited!!
Please keep us in your prayers ���
Translation Caption :
37 minggu! � Aplikasi saya
mengatakan bayinya sekitar 6,5 lb dan
seukuran melon kecil! Baby sedang
berlatih gerakan karatenya sementara
355
 saya praktis berdiri di atas kepala saya
mencoba membuatnya berubah menjadi
posisi kepala di bawah �� Tas rumah
sakit saya penuh sesak dan saya punya
semua perlengkapan dan tidak tahu
tapi kami sangat bersemangat! ! Tolong
jaga kami dalam doa-doamu ���
In this data, four raters gave three
points, sixteen raters gave two points and
none of the raters gave one point. The
number of raters who gave two points (16
raters) shows that the translation is quite
natural but there are still clauses, phrases
and sentences that are less familiar to
readers.
Unacceptable
Score one is for translation that the
technical terms are not commonly used and
not familiar to the readers; phrases, clauses,
and sentences are not in accordance with
the rules of the target language.
Instagram Machine Translation
Instagram Machine Translation has a
percentage of 25.88% for data that is
considered unacceptable. Below is an
example of an unacceptable translation
from Instagram Machine Translation:
(Datum A-1)
Original Caption :
Dressing room/Office progress! I asked
you guys what I should put on the other
side of my vanity to balance out the
mirror and several of you are saying a
plant, but I’m not sure if I have enough
space for a plant there!! I was thinking
some floating shelves for more shoe
storage? I don’t want it to look too
cluttered � Let me know what you
think?! linked
decor here!
http://liketk.it/2SqVI #liketkit @liketok
now.it
Translation Caption : Furthermore, for translations that get 2
Kemajuan ruang ganti/kantor! Saya
bertanya kepada kalian apa yang harus
saya letakkan di sisi lain kesombongan
saya untuk menyeimbangkan cermin
JELLT (Journal of English Language and Literature Teaching)
Vol. 7, No. 2 November 2022
P.ISSN : 2548-7728 E.ISSN : 2599-0373
dan beberapa dari kalian mengatakan
tanaman, tetapi saya tidak yakin apakah
saya memiliki cukup ruang untuk
tanaman di sana!! Saya berpikir
beberapa rak terapung untuk
penyimpanan sepatu lebih banyak?
Saya tidak ingin ini terlihat terlalu
kacau � Beri tahu saya apa yang Anda
pikirkan?! Dekorasi yang ditautkan di
sini! http://liketk.it/2SqVI #liketkit
@liketoknow.it
In the translation, there were thirteen
raters who gave one point. This shows that
the translation is in the unacceptable
category. Which means that the meaning of
the translation cannot be understood by the
reader at all.
Google Translate
In this category none of the data gets
one point from more than ten appraisers.
This indicates that Google Translate
produces acceptable translations.
Readability of the Translation
There are three levels for this category,
namely high level readability, medium level
readability and low level readability. The
translation results that fall into the category
of high level readability are translations in
which the words, technical terms, phrases,
clauses, sentences, or text translations can
be easily understood by readers. Then the
medium-level readability is for translations
that make the reader have to read the
translation more than once to be able to
understand its meaning. The last one, Low-
level readability, is for translation whose
meaning cannot be understood at all by the
reader. In this category, Google Translate is
still superior to Instagram Machine
Translation because it has a percentage of
the amount of data that has 3 points as
much as 46.25%, while Instagram Machine
Translation only has a percentage of 29.5%.
points, Instagram Machine Translation has
a higher percentage of 54.5% while Google
Translate has a percentage of 45.75%. Then
for translations that get 1 point or fall into
356
the unreadable category, Instagram sangat bersemangat! Mereka bahkan
Machine Translation has a percentage of belum bertemu Leo! �
16% while Google Translate has a In the translation above, there are eleven
percentage of 8%. raters who agree that the translation has a
Table 3. The amount of data and high level of readability. A high level of
percentage on the aspect of acceptability readability means that words, technical
terms, phrases, clauses, sentences, or text
Quality translations can be easily understood by the
Readability reader.
Translation
Machine

No Google Translate
(High Level
Readability

Readability

Readability
(Low Level
(Medium

. Google Translate has a percentage of

Level
3

2)

1)

)
46.25% for data that is considered to have a
high level of readability. Below is an
example of data that falls into that category:
(29.5%)

(54.5%)

(Datum B-3)
(16%)
IMT

1.
236

436

128

Original Caption :
We’ve been very busy ��� having the
Total Data 800 Data
best time on our family vacation!! We
are staying at the most
(46.25%)

(45.75%)

beautiful @vrbo home and is seriously

(8%)

2. GT
370

366

64

perfect for our big family! We have our

High Level Readability
A translation is considered to have a
high level of readability if the words,
technical terms, phrases, clauses, sentences,
or text translations can be easily understood
by the reader.
Instagram Machine Translations
Instagram Machine Translation has a
percentage of 29.5% for data that is
considered to have a high level of
readability. Below is an example of data
that falls into that category:
(Datum A-27)
Original Caption :
Happy Friday and happy March
Lovelies!! Let’s have a coffee date to
celebrate � � My sisters are coming to
town this weekend and I’m so excited!
They haven’t even met Leo yet! �
Translation Caption : to read it more than once to understand its
Selamat hari Jumat dan selamat Maret
Lovelies!! Mari kita minum kopi untuk
merayakanya � � Saudariku akan
datang ke kota akhir pekan ini dan aku
JELLT (Journal of English Language and Literature Teaching)
Vol. 7, No. 2 November 2022
P.ISSN : 2548-7728 E.ISSN : 2599-0373
own private pool with this stunning
view!! �#ad
Translation Caption :
Kami sangat sibuk ��� bersenang-
senang di liburan keluarga kami!! Kami
menginap di rumah @vrbo yang paling
indah dan sangat cocok untuk keluarga
besar kami! Kami memiliki kolam
renang pribadi dengan pemandangan
yang menakjubkan ini!! �#iklan
In the translation above, there are fifteen
raters who agree that the translation has a
high level of readability. A high level of
readability means that words, technical
terms, phrases, clauses, sentences, or text
translations can be easily understood by the
reader.
Medium Level Readability
A translation is considered to have a
medium level of readability if the reader has
meaning.
Instagram Machine Translation
Instagram Machine Translation has a
percentage of 54.5% for translations that
have a medium readability level. Below is
357
an example of data that falls into that reader has to read it more than once to
category: understand its meaning.
(Datum A-11) Low Level Readability
Original Caption : A translation that is categorized as
Bringing out all the cute Middle Eastern having a low level of readability is a
style photos for Ramadan � I hope you translation whose meaning cannot be
girls are making the most out of these understood at all. On both machine
last few beautiful nights!! ��� translators, there is no data that gets one
point from more than ten raters. This shows
Translation Caption :
that both Google Translate and Instagram
Membawa semua foto gaya Timur Machine Translation do not produce
Tengah yang lucu untuk Ramadhan � completely unreadable translations.
Saya harap kalian membuat yang
terbaik dari beberapa malam yang Similarities and Differences between
indah ini!! ��� Google Translate and Instagram
Machine Translation
In the translation above, there are
fourteen raters who agree that the Original Caption
translation has a medium level of Alaina wearing the first outfit I ever
readability. A medium level of readability bought her when I found out I was having a
means that the reader has to read it more girl!!! Swipe to see when I bought the outfit
than once to understand its meaning. and see last photo for reality of this
photoshoot �
Google Translate
Google Instagram
Google Translate has a percentage of Translate Machine Translation
45.75% in translations that have a low Alaina Alaina
readability level. Below is an example of mengenakan mengenakan pakaian
data that falls into that category: pakaian pertama pertama yang pernah
(Datum B-24) yang saya beli saya beli ketika saya
ketika saya tahu saya punya anak
Original Caption : mengetahui bahwa perempuan!!! Swipe
Who else was actually relieved when saya memiliki untuk melihat pakaian
seorang gadis!!! dan melihat foto
instagram went down yesterday?! At Geser untuk terakhir untuk realita
first I thought I got hacked � after I pemotretan ini �
melihat saat
realized I wasn’t, It felt like a mini saya membeli
vacation! Makes me realize I need to pakaian tersebut dan
unplug way more often. What about lihat foto terakhir
you? untuk realita
pemotretan ini �
Translation Caption :
Siapa lagi yang sebenarnya lega ketika
instagram down kemarin?! Awalnya Similarities and Differences
saya pikir saya diretas � setelah saya In the translation results, there is a
sadar ternyata tidak, Rasanya seperti difference in the word "pernah", where in
liburan mini! Membuat saya sadar the translation produced by Instagram
bahwa saya perlu mencabut lebih Machine Translation the word "ever" is
sering. Bagaimana denganmu? translated into "pernah" while in Google
In the translation above, there are Translate the word is omitted. Another
thirteen raters who agree that the translation difference is in the word "swipe" which
has a medium level of readability. A translated to "geser" in Google Translate
medium level of readability means that the while on Instagram Machine Translation the
word is still written "swipe". In addition,
JELLT (Journal of English Language and Literature Teaching) 358
Vol. 7, No. 2 November 2022
P.ISSN : 2548-7728 E.ISSN : 2599-0373
the sentence "Swipe to see when I bought Mereka bahkan �
the outfit" translated properly by Google belum bertemu Leo!
Translate into "Geser untuk melihat saat �
saya membeli pakaian tersebut," while on Similarities and Differences
Instagram Machine Translation, the word In the translation, the word "coffee
"bought" is omitted and only translated as date" translated to "kencan minum kopi" in
"Swipe untuk melihat pakaian" Google Translate, while in Instagram
Original Caption Machine Translation, the word "date" is
Feeling so so grateful ��� I cannot omitted and only translated to "minum
believe we only have a few more weeks kopi". Furthermore, the word "My sister"
until we are a little family! God willing � translated to "kakak-kakaku" on Google
Translate while on Instagram Machine
Translation, the word translated to
Google Instagram
Translate Machine Translation "saudariku"
Merasa sangat Merasa sangat CONCLUSION
bersyukur ��� bersyukur ��� Aku The results of this study indicate that the
Saya tidak percaya tidak percaya kita
kita hanya memiliki hanya memiliki
two machine translators, namely Google
beberapa minggu beberapa minggu lagi Translate and Instagram Machine
lagi sampai kita sampai kita menjadi Translation, both produced less accurate,
menjadi keluarga keluarga less acceptable, and less readable
kecil! Insya kecil! InsyaAllah translations, but Google Translate is
Allah � � considered better because it has a higher
percentage of translations that get 3 points
Similarities and Differences in all aspects. In the aspect of accuracy, the
In this translation, almost both machine data that is considered accurate or gets 3
translations produce the same translation. points on google translate is 32.5% while
The only difference is that "I" translated to the Instagram machine translation is only
"Saya" in Google Translate, while on 15%. Then for data that is considered less
Instagram Machine Translation, the word accurate or gets 2 points, google translate
"I" translated to "Aku." has a percentage of 58.33% while Instagram
machine translation has a higher percentage
Original Caption of 61.67% and for data that is considered
Happy Friday and happy March inaccurate or gets 1 point, Google translate
Lovelies!! Let’s have a coffee date to has a small percentage, namely only 9.17%
celebrate ��� My sisters arecoming to and Instagram Machine translation has a
town this weekend and I’m so excited! percentage of 23.33%.
They haven’t even met Leo yet! � For the acceptability aspect, Google
Translate has a percentage of data that gets
Google Instagram 3 points as much as 48% while Instagram
Translate Machine Machine Translation, only has a percentage
Translation
of 22.62%. Furthermore, for data that is
Selamat hari Mari kita considered less acceptable or gets 2 points,
Jumat dan selamat minum kopi untuk
Maret Lovelies!! merayakanya ���
google translate has a percentage of 44.5%
Mari kita kencan Saudariku akan while Instagram Machine Translation has a
minum kopi untuk datang ke kota akhir higher percentage of 51.5%. Then, for data
merayakannya ��� pekan ini dan aku that is considered unacceptable or gets 1
Kakak-kakakku akan sangat bersemangat! point, google translate has a small
datang ke kota akhir Mereka bahkan
pekan ini dan aku belum bertemu Leo!
percentage of only 7.5% while Instagram
sangat bersemangat! Machine Translation has a much higher
JELLT (Journal of English Language and Literature Teaching) 359
Vol. 7, No. 2 November 2022
P.ISSN : 2548-7728 E.ISSN : 2599-0373
percentage of 25.88%. Kualitas Terjemahan. Kajian Linguistik
For the readability aspect, the data that dan Sastra, Vol. 24, 39-57.
is considered to have a high readability Peng, H. (2018, March). The impact of
level or get 3 points on google translate is machine translation and computer-aided
46.25% while on Instagram Machine translation on translators. In IOP
Translation there is only 29.5%. Conference Series: Materials Science
Furthermore, for data that is considered to and Engineering (Vol. 322, No. 5, p.
have a medium level of readability, google 052024). IOP Publishing.
translate has a percentage of 45.75% while Macken, L., Prou, D., & Tezcan, A. (2020,
Instagram machine translation has a June). Quantifying the effect of machine
percentage of 54.5%. Then, for data that is translation in a high-quality human
considered to have a low readability level translation production process.
or gets one point, Google Translate has a In Informatics (Vol. 7, No. 2, p. 12).
percentage of 8% and Instagram machine Multidisciplinary Digital Publishing
translation has a percentage of 16%. Institute.
Meilasari, P. (2019). Instagram Translation
It is recommended that the use of
Machine: Widya Warta, 124-133.
machine translators is accompanied by
Ninggolan, W., Sinurat, B., & Hutauruk, B.
human intervention to produce translations
S. (2020). Accuracy of the Translations
that are accurate, acceptable, and have a
of English Noun Phrases into
high level of readability. For future
Indonesian Using Google Translate.
researchers, it is hoped that they will be
Journal of English Teaching as a
able to conduct more in-depth research by
Foreign Language, 66-87.
conducting interviews with assessors to
Thriveni, C. (2002). Cultural elements in
find out the reasons why they give points
translation: The Indian perspective.
one, two, or three to a translation result.
Translation Journal. 6/1
Apart from that, they can also research
Tripathi, S., & Sarkhel, J. K. (2010).
other machine translations such as
Approaches to machine translation.
SYSTRAN, Microsoft Translation,
Annals of Library and Information
iTranslate, Translate Dict, Yandex
Studies, 388-393.
Translate, and Microsoft Bing.
REFERENCES
Ahrenberg, L. (2017). Comparing machine
translation and human translation: A
case study. In RANLP 2017: The First
Workshop on Human-Informed
Translation and Interpreting
Technology (HiT-IT) (pp. 21-28).
Association for Computational
Linguistics.
Halimah. (2018). Comparison Of Human
Translation With Google . Jurnal
Pendidikan Bahasa dan Sastra, 11-29.
(https://www.ef.com/assetscdn/WIBIwq6R
dJvcD9bc8RMd/cefcom-epi-
site/reports/2021/ef-epi-2021-
english.pdf, 2021)
Nababan, M., Nuraeni, A., & Sumardiono.
(2012). Pengembangan Model Penilaian

JELLT (Journal of English Language and Literature Teaching) 360

Vol. 7, No. 2 November 2022
P.ISSN : 2548-7728 E.ISSN : 2599-0373